SPLIT-SPECTRUM PROCESSING FOR FLAW-TO-GRAIN ECHO ENHANCEMENT IN ULTRASONIC DETECTION

NIHAT MUSTAFA BILGUTAY, Purdue University

Abstract

When the range cell of an ultrasound system contains many unresolved and random reflectors such as grains, the overlapping echoes which result make the detection of flaws within the range cell difficult even when the flaws are substantially larger than the grains. This problem arises in the ultrasonic examination of many industrially important materials such as titanium, ceramics and stainless steel welds. The non-random nature of grain echoes prevents the utilization of time averaging or correlation techniques which are effective in random noise suppression. Since the grains are small and closely spaced, their echoes result in a complicated interference pattern which is significantly more sensitive to shifts in the transmitted frequency compared to flaws which are much larger in size. Therefore, decorrelation in grain echoes can be obtained by shifting the transmitted frequency. This basic principle has led to the development of a split-spectrum processing technique for an ultrasonic flaw detection system which improves the flaw-to-grain echo ratio in large grained materials. The enhancement is achieved by partitioning a wideband received spectrum to obtain frequency shifted bands, which are then processed to suppress the grain echoes with respect to the flaw echo using a novel minimization and conventional averaging algorithms. Simulation was performed using a set of frequency shifted signals to examine the effectiveness of each algorithm. The results confirm the feasibility of grain noise suppression by frequency decorrelation techniques and show the minimization algorithm to provide enhancement in flaw visibility which is superior to the averaging algorithms. An experiment system was developed which obtains frequency shifted signals from the wideband received signal by digital filtering. Titanium and stainless steel samples were examined to evaluate the performance of the experimental system. The minimization algorithm was seen to provide significant flaw visibility enhancement with respect to the original data in contrast to the limited improvement observed for the averaging algorithms. These experimental results are shown to be in close agreement with the simulation results. The behavior of the minimization algorithm was evaluated qualitatively based on theoretical range cell models for both deterministic and stochastic scatterers. Based on these models, the dependence of the algorithm performance on the transmitted signal bandwidth and frequnency spacing is described. The effect of flaw amplitude on the performance of the averaging and minimization algorithms was examined. The experimental results show that the minimization algorithm is capable of recovering a flaw echo which is smaller than the grain echoes even when the averaging algorithms cannot. The effect of incidence angle on the performance of the minimization algorithm was also examined but shows no unusual sensitivity. Finally, the examination of stainless steel samples which result in Rayleigh and Stochastic scattering show that the minimization algorithm can achieve enhancement even when multiple scattering effects are present. These results also indicate that the performance of the minimization algorithm depends on grain size, which generally, deteriorates as the grain size increases. The overall experimental results suggest that the minimization algorithm has potentially useful industrial applications in the examination of large grained materials such as titanium, ceramics and weld components, where improved flaw detection is highly desirable.

Degree

Ph.D.

Subject Area

Electrical engineering

COinS